Slow Incident Response, Serious Data Exposure.

A recent decision by the UK Information Commissioner’s Office (ICO) highlights the importance of having effective cyber security controls and a timely incident response process. Capita plc and Capita Pension Solutions Limited were fined a combined £14 million after a cyber attack in March 2023 led to the theft of personal data relating to more than 6 million individuals. [1]

What happened?

  • Attackers gained access to Capita’s systems after a malicious file was downloaded onto an employee’s device.

  • Although the incident triggered an alert, the affected device was not isolated for 58 hours.

  • During that period, the attackers were able to move through Capita’s systems and access large volumes of personal data.

  • The compromised data included information relating to pension records, staff records, financial data and special category data.

  • The ICO fined Capita plc £8 million and Capita Pension Solutions Limited £6 million, bringing the total penalty to £14 million.

Why was this a problem?

  • Capita failed to respond appropriately to security alerts, despite the seriousness of the risk.

  • The delay in containing the affected device allowed the attackers to continue accessing systems for more than two days.

  • The ICO found weaknesses in Capita’s security arrangements, including inadequate controls to prevent privilege escalation and unauthorised lateral movement.

  • Capita also failed to ensure that its risk assessments and penetration testing were sufficient in the context of the scale and sensitivity of the data it processed.

  • These failures fell short of the requirements under Articles 5(1)(f) and 32 of the UK GDPR.

What does it mean for your organisation?

This case shows that detecting a cyber incident is only part of the obligation. Organisations must also be able to act quickly, contain the threat, and limit the impact on individuals. Where large volumes of sensitive data are involved, regulators will expect security monitoring, escalation procedures and incident response arrangements to work effectively in practice.

To avoid similar pitfalls, organisations must:

  • Respond promptly to security alerts: alerts must be reviewed, escalated and acted upon within defined timeframes.

  • Test incident response procedures: organisations should run practical exercises to ensure teams know how to contain threats quickly.

  • Limit attacker movement: access controls, privilege management and network segmentation should reduce the risk of lateral movement.

  • Review penetration testing and risk assessments: testing should reflect real risks, business complexity and the sensitivity of the data processed.

  • Ensure adequate resourcing: security teams must have the capacity and authority to respond effectively when incidents occur.

The takeaway? It is not enough to have cyber security tools in place if the organisation cannot respond when those tools identify a threat. Regulators are focusing closely on whether security controls work operationally, not just whether they exist on paper.

Tales from the coalface

Each month, we examine Data Protection issues encountered by our clients. This month, we focus on the growing need for clear internal policies on AI use, and why organisations should not wait until something goes wrong before setting boundaries.

AI tools are now appearing across everyday business processes, from drafting emails and summarising meetings to analysing documents, generating code, preparing marketing content and supporting customer service. While these tools can improve efficiency, they also create significant data protection risks when staff use them without clear guidance, approval or oversight.

When AI Use Becomes a Compliance Blind Spot

Across sectors, we are seeing a recurring issue: employees are experimenting with AI tools before the organisation has decided what is, and is not, permitted. In many cases, this is not done with bad intentions. Staff may simply be trying to work faster or improve outputs. However, where personal data, confidential client information or business-sensitive material is entered into public or unapproved AI tools, the organisation may quickly lose control over that data.

What’s the risk of unclear AI use?

  • Loss of control over personal data: Once information is entered into an AI tool, it may be difficult to determine where it is stored, how long it is retained, who can access it, or whether it is used to improve the model.

  • Unclear controller and processor arrangements: If the organisation has not reviewed the AI provider’s terms, it may not know whether the provider acts as a processor, independent controller or another party.

  • International transfer risks: Some AI tools may process or store data outside the UK or EEA, creating transfer issues if appropriate safeguards have not been assessed.

  • Confidentiality concerns: Staff may accidentally disclose employee, client, patient or customer data into tools that have not been approved for such use.

  • Poor alignment with GDPR principles: Uncontrolled AI use can undermine data minimisation, transparency, purpose limitation, security and accountability.

Why it happens:

AI use often spreads informally before governance catches up. The tools are easy to access, low-cost or free, and often appear harmless at first glance.

  • Staff are not sure which AI tools are approved.

  • There is no clear policy explaining what data can and cannot be entered into AI systems.

  • Procurement and IT teams are not always involved before tools are adopted.

  • DPIAs and vendor assessments are skipped because the tool is seen as “just a productivity aid”.

  • Training has not kept pace with the speed of AI adoption.

How to fix it:

The key is not to ban AI entirely, but to create clear, practical rules that allow responsible use while protecting personal data.

  • Implement an AI use policy: Define approved tools, prohibited uses, permitted data types and escalation routes.

  • Classify data before use: Make clear that special category data, client confidential data and employee records must not be entered into unapproved tools.

  • Review AI suppliers properly: Assess data protection terms, retention, security, sub-processors and international transfers.

  • Require DPIAs where needed: Especially where AI is used for monitoring, profiling, automated decision-making or high-risk processing.

  • Train staff regularly: Employees should understand both the benefits of AI and the risks of careless use.

AI policies are no longer a “nice to have”. They are a practical necessity for any organisation whose staff are already using, or likely to use, AI tools in their daily work.

Without clear rules, AI adoption can become fragmented, undocumented and difficult to control. Organisations should treat AI use as a governance issue, not just a technology trend. A good policy helps staff innovate safely, gives managers confidence, and supports compliance with data protection obligations.

If your organisation needs help preparing an AI use policy, reviewing AI tools, or updating your data protection framework for AI-related risks, Ambit Compliance can help.

Do not hesitate to contact us at info@ambitcompliance.ie and check further about us on our website https://www.ambitcompliance.ie/.

Contributors to this blog: Gillian Traynor and Facundo Heredia.

References

[1] https://ico.org.uk/about-the-ico/media-centre/news-and-blogs/2025/10/capita-fined-14m-for-data-breach-affecting-over-6m-people/

Previous
Previous

What Caught Our Eye This Week?

Next
Next

What Caught Our Eye This Month? (Part 2)